Performance Analysis And Study Of The Campus Network Based On Flow Characteristics

As the Internet continues to expand and applications continue to emerge, demands for online monitoring and management network traffic become urgent. Therefore, the study on characteristics of network traffic is important to network management, planning and development. Network traffic measurement can analysis network traffic characteristics, acquire network status, and achieve fault location and recovery based on detailed traffic information. The user application intensity, frequency, load and other behavioral models are also be obtained.Based on deep analysis and discussion on the campus network traffic characteristics, design and implementation of network traffic monitoring and analysis system is provided. It also proposes HHT improved algorithms for network anomaly detection in the identification of abnormal traffic based on self-similar theory.The main work of this paper include the following:First of all, the monitor network traffic system based on the MIB object is deployed, which realize the entire campus network traffic monitoring, early warning network traffic, network equipment log management, network equipment status monitoring and etc. It is also provided to do a deep cyclical analysis to the traffic of the whole campus, and analysed self-similarity of network traffic from different levels.Secondly, after capturing the campus network outbound traffic we do a deep research and analysis on distribution characteristic of data flow, and analysis network traffic from the transport layer and application layer protocol, so that we can understand composition of network applications and status of network conveniently.Finally, because campus network is influenced by the characteristics of a single point, detecting abnormal traffic by calculating self-similarity directly does not work well. Based on the theory that traffic trends can be broken down into component items and other random items, this paper proposes an improved anomaly detection algorithm, which use empirical mode decomposition to eliminate the trend of network traffic. Some experiments on Dong Hua campus network have been done, the results show that the improved detection of abnormal traffic algorithm improve the accuracy of anomaly detection to a certain extent.